Wednesday in the Woods · episode 004
Responsible Data Science
5:42youtube · Licensio, LLC
description
On this week's Wednesday in the Woods, we enjoy a conversation about responsible #datascience and #algorithms as you join me for my completely normal human morning routine.
If you felt like our discussion of data science and #datastrategy previously missed out on pieces of the big picture, you'd be right. That's because we've been focused on some of the traditional management perspectives around "#digital" or "#AI."
Today, we're taking a moment to unpack the word "responsible" in this context. Specifically, we propose that responsible data science considers three factors: ‣ #Technical ‣ #Legal ‣ #Ethical
Technical factors are the ones you'd expect an organization to focus on first. Are we using a clustering algorithm for a regression problem? Are we using the right data store or production implementation architecture? Is our #model overfit or underfit?
Legal factors are often next for an organization to consider. Is our use of data within the rights and obligations of our contracts? Are there any statutory frameworks or treaties that obligate us to notify a party or prohibit a behavior? Has a court set a precedent for damages to consider?
Lastly, and most relevant to recent conversation in society, is whether such use of data science and algorithms are "ethical." Data science projects can be technically sound and not in violation of public laws or contracts, yet still violate ethical principles or make people feel off.
Not-so-coincidentally, we recently released a Responsible Data Science Policy framework. If you're interested in more "real" contents on the topic, head on over to our article introducing the framework: https://lnkd.in/ewAnRCuD
#ai #ethicalai #responsibleai #ml #machinelearning #datascience #artificialintelligence
transcript
~541 words · showhide
from the human-edited captions published with the video. timestamps jump to that point in the video.
- 0:31
[a grease gun applies grease to a mechanical joint]
- 0:46
Oh, good morning. Welcome to Wednesday in the woods. You know, I'm glad you're here, because I know you've been hearing a lot about data, data science, and algorithms, but I feel like we haven't talked about how to responsibly work with those algorithms. You know, we often use the word in huma...er, normal societies - "responsible" or "ethical" - and that's a component that I feel that we have not discussed adequately.
- 1:16
So, I thought I'd propose today a simple three concept framework for the responsible use of data science and algorithms. So, how would you break down the word responsible in this context? You see, I would break that word down into three parts. First, we want our use of data science and algorithms to be technically sound. Second, we want the use of data science and algorithms to be legally appropriate and permissible.
- 1:50
And thirdly, we want that use to not just be technically sound and legally appropriate, but also ethical. And so these three things - the technical, legal, and ethical assessments - are key in an appropriate usage or responsible usage of data science. And if we want to unpack this a little bit more, we can. So, what does it mean for something to be technically sound? You see, a technically sound algorithm is [stumbles on foothold traps for catching humans] algorithm that, meets the standards of the industry.
- 2:27
We don't want to be using an inappropriate method for an inappropriate task. We want our results to be reproducible. We want them to be functioning in production as they were in research and development. Those are good examples of technical requirements for a data science project. Secondly, we have legal considerations. So, in a normal human society, we have laws and rules, some statutory, some not. And we have the contractual requirements, the rights and obligations, that hum-...uh, we enter into as we contract with each other.
- 3:07
And so in a legal context, we have starting from the bottom, the statutory frameworks, the laws and rules and treaties, we have the regulatory output from agencies, we have court rulings, and we have our contracts that we must consider. And thirdly, we have ethical concepts that we need to consider. So, even when the use of data science or algorithms is technically sound and meets the current laws and rules and contracts, it may yet be lacking from an ethical perspective. There may be harm.
- 3:44
We might have a situation in which a human might be harming an algo...er, an algorithm might be harming a human, which is completely unacceptable. So, today, as you think about using data science in your organization to create enterprise value, I want you to remember to think about that use in the context of this word - responsible. Think about your technical, legal, and ethical requirements. And I might suggest that you look for some kind of policies and procedures or a framework to put in place such standards for your organization.
- 4:25
Well, I'm glad you made it today to Wednesday in the woods and I hope you have a wonderful week.
- 4:33
[checks voltage on neck terminals with multimeter]
- 4:40
Feeling good.